CatNetVis: Semantic Visual Exploration of Categorical High-Dimensional Data with Force-Directed Graph Layouts

Abstract

We introduce CatNetVis, a novel method of representing semantical relations in categorical high-dimensional data. Traditional methods provide insights into many aspects of visual exploration of data. However, most of them lack information on relations in between categories or even clusters of categories. The force-directed network layout utilized by CatNetVis enables a lightweight approach in order to explore such semantical relations. The connections within the network are perceived as an intuitive metaphor for clusters of connections/relations in categorical data denoted as communities. While the user interacts, visual encodings such as information about the entropy and frequencies allow a fast perception of relation between categories and its frequencies, respectively. We illustrate how CatNetVis performs as an effective addition to traditional methods by demonstrating the method on an example data sets and comparing it to conventional methods.